This invention relates to performance management of a computer system which includes a storage system, and more particularly, to a method of monitoring performance of a computer system by using an agent.
Conventionally, a storage network in which a plurality of computers access an integrated storage system via a network is realized. The storage network as described above can increase utilization efficiency of storage systems whose sizes are on the increase, and reduce management costs. Thus, the storage network is being put into wide use as an architecture for a data center. In a computer system into which the storage network as described above is introduced, to monitor and adjust performance of each application system, it is necessary to grasp a correlation among hardware equipment or the like and time transition thereof by comprehensively collecting pieces of performance information regarding various hardware and software equipments constituting the network. This is because the application systems before the introduction of the storage network have been independent of each other in a unit of a server composed of a processor unit and an external storage system connected to the processor unit, so it has been unnecessary to consider performance interferences between the application systems. However, in the storage network environment, there is a possibility that performance interference is caused between the application systems in a part (e.g., network system or storage system) shared by the plurality of computers.
An example of a performance monitoring technique in a networked computer system is disclosed in U.S. Pat. No. 6,505,248. According U.S. Pat. No. 6,505,248, performance is managed by an agent arranged for each of hardware resource and software resource which are monitoring targets, and management software for uniformly managing performance information of the entire system.
Each of the agents obtains the performance information by directly communicating with the monitoring target. Specifically, the agent executes polling with respect to the monitoring target to obtain a metrics value of resources. Here, the resources are physical or logical elements constituting the computer system. The metrics is a performance monitoring item (e.g., CPU utilization rate or number of I/Os) constituting resource performance information.
On the other hand, the management software collects and stores the pieces of performance information obtained by the agents, and provides the stored pieces of performance information according to a request from an administrator or the like. Accordingly, it is possible to uniformly manage operation states and performance of the hardware resource and the software resource constituting the computer system.
When the size of the uniformly managed computer system increases, the number of resources to be monitored also increases. Along with an increase in the number of resources, an amount of the performance information which the management software must obtain also increases, whereby it is necessary to process a large amount of performance information. JP 2005-157933 A discloses a method of automatically adjusting a target range from which performance information is to be subsequently obtained and an obtaining frequency based on the information obtained by the management software.